Smart Model Design for Recommending the Optimal Concentration of F. golbaniflua Root Extract in PLA to Achieve Tunable Fibrous Properties

کد مقاله : 1179-ICOC

نویسندگان

فاطمه دارائی *، طیبه رمضانی فرزین دانشگاه سمنان

چکیده مقاله

Subject and objectives: Controlling the structural and tunable textural properties of electrospun scaffolds remains a major challenge in skin tissue engineering [1]. Variations in the concentration of bioactive additives, such as Ferula golbaniflua root extract, can significantly influence fiber diameter; however, these effects are generally nonlinear and dependent on multiple processing parameters. In this study, an intelligent predictive model based on multiple machine learning algorithms was developed to identify the optimal concentration of extract in a PLA solution, aiming to fabricate fibers resembling the architecture of the skin extracellular matrix.    Methods: A polymer solution was prepared by dissolving PLA in chloroform (7.5% (w/v)). After complete homogenization, F. golbaniflua root extract was added at concentrations of 0%, 1%, and 2% (w/v). Electrospinning was performed under constant conditions: a 15 cm tip-to-collector distance and an applied voltage of 13 kV. Fiber mats were collected on glass slides. Optical microscopy was used to image the scaffolds, and fiber diameters were measured using ImageJ software. For each extract concentration, ten independent measurements were collected to ensure statistical reliability. To analyze the relationship between extract concentration and fiber diameter, and to predict diameters at untested concentrations, three machine learning regression algorithms were employed: second-degree polynomial regression, support vector regression (SVR), and Gaussian process regression (GPR). Experimental data were used to train the models and determine the extract concentration required to achieve the target fiber diameter of approximately 300 nm.    Results: Comparison of the three models indicated that GPR provided the highest predictive accuracy. The optimal extract concentration predicted by GPR was 0.786%, with an uncertainty of ±12.8 nm. Experimental observations confirmed that increasing the extract concentration led to a reduction in fiber diameter. The predicted optimal concentration falls within a region where electrospinning behavior is stable and the fiber diameter shows a nearly linear dependence on concentration. Concentrations above 2% produced excessively thin fibers (<200 nm), which may be unsuitable for skin tissue engineering. The optimal range therefore ensures consistent fiber morphology and an average diameter closely matching the ECM architecture (~300 nm).    Conclusions: Overall, this study demonstrates that combining experimental data with machine learning analysis provides a reliable and effective approach for designing bioactive electrospun scaffolds and optimizing their formulation. The GPR-based model enables accurate prediction of fiber characteristics, even with limited experimental data, and supports rational design strategies for tissue engineering applications.

کلیدواژه ها

Tissue engineering, PLA, F. golbaniflua, Electrospinning, Machine learning

وضعیت: پذیرفته شده